Anthropic's $1.75 Trillion IPO: Are Investors Overreacting to the Risks?

0:00This episode is brought to you by Indeed. You're ready to level up your business, but the pressure to find the right person to make that happen is intense. Indeed Sponsored Jobs can help carry the load. It boosts your job post to help it reach quality talent fast. Cut through the chaos with Indeed Sponsored Jobs. Get a$75 sponsored job credit at indeed.com slash podcast. Terms and conditions apply. Hiring now? Then this is a job for Indeed Sponsored Jobs. Anthropic has officially filed its S1 prospectus, setting up a historic public debut targeting between$1.5 and$1.75 trillion. But here is the number that should give everyone a pause.

0:42The risk factors section alone spans 94 pages and more than 60 ,000 words, disclosing catastrophic safety liabilities, an$80 billion headline loss, and the governance model where outside shareholders will have no real say in the boardroom. So today, we are unpacking what's actually in the filing to answer the central question, is ticker ant? Great ticker, by the way. How do you feel about that? What? Well, I don't want to be responsible for an existential risk to humanity.

1:16Piers Curran: well this is it i mean like when you're you know that's going to ruin your plans for you know launching the when your public persona goes oh the anthony channel media company that's right i did have it i thought i was going to do ant but maybe maybe the media company will be bigger than anthropic and i can just take it off them who knows yeah so we'll look at then in this conversation appears whether trying to get to and we definitely want people's participation in the comments is this a defining enterprise investment of the next decade or an unprecedented governance trap because there's some super interesting things certainly within he tucked within this s1 prospectus if you've had the time to read through it that is which i know that you have piss but yeah absolutely yeah every every word what was it 300 and how many pages was it over 300 pages, wasn't it?

2:15Piers Curran: Yeah, unusually long. Yes, I have not read the whole thing. Just to get that out up front. But look, I guess it's kind of, obviously it's a major, just we'll get into the detail, but stepping back, obviously it's a major event, clearly. It's going to break the record. A bit of a shame for Elon, maybe, given that he broke the record for the biggest IPO in history a couple of months ago and his record is going to be taken off him almost immediately. So obviously it's a major moment for equity markets and the chronicle of the equity market journey over the centuries. This is definitely a major event.

3:04Piers Curran: It's about time. Obviously we've been talking about this for months, months, months, months. So it's nice to just get it on. Obviously, there was some doubts. Is it going to happen? Is it not? We've been through a bit of a wobble. The AI bubble, if it is one, has deflated somewhat over the last few weeks. Certainly, software, or I should say semiconductor stocks have taken a bit of a battering over the last few months, really. So it's nice to now just get this going. Actually, we'll talk about maybe broader market reactions later, but it's been a very positive one from a broader market sort of perspective.

3:49Piers Curran: So some positive news, I would say, for the AI boom trade. And that's because some of the details in this, if you can get past the alarmist headlines and the 90 pages of, you know, we're going to break the whole world, if you can get past that, there's actually some good stuff from there. Yeah. So let's look at that. And we've tried to break this down into really some standout sections because the bits that really stood out were things like the five share class structure. So some of the governance architecture, there's also the financials. So what does it actually look like now? We get to see a little bit more under the bonnet of how they're performing and their margins and so forth, compute power.

4:33And then also this idea of the concentration risk on their client mix. So let's get into some of this. So maybe we could start with the governance side, because when I saw this drop, I was like, wow, five different classes of share, of which I didn't quite initially get until I started reading it in more detail. But so what do we have here?

4:52Piers Curran: It's complicated. I don't know. The purist out there is, well, look, you're an equity holder. You own a share of the company. And the purist would say, well, if you own a share, well then proportionately you should have a say in how that company that you own is run right so the purist will be um very unhappy about this complex setup but it's you know none of this is well maybe some of this is new but it's it's kind of not it's not like it's breaking the mold that much from some of the stuff that the likes of meta and alphabet have done down the years right where the founders try and retain control.

5:31Piers Curran: But look, they think they're going to incorporate in Delaware, which is where everyone incorporates, unless you're Elon Musk, who did incorporate in Delaware and has changed his mind and gone off to Texas because Delaware courts blocked his massive incentive structure. But anyway, they're incorporating in Delaware as a PBC. And basically, this basically legally binds the board of directors to balance three distinct elements. Okay, so something called stockholder pecuniary interests. You've got the best interests of materially affected parties. And then number three, which is the interesting one, is the specific public benefit mandate.

6:18Piers Curran: So basically, this is their responsible development and maintenance of advanced AI. for the long-term benefit of humanity, right? So they've kind of got this core moral, what they feel like it was a core moral obligation to the human race to deliver this, you know, once in a multi-generation kind of moment to deliver this new technology responsibly, as I said, without breaking the planet. Haven't we heard this before though? Yeah. Yeah. And what, Sam Altman was all for this until he then thought, oh, I can make a bobble or two here. Yeah, this is it. I think with Sam Altman and OpenAI, it was a bit different in that they were a non-profit organization.

7:07Piers Curran: That's how they were set up. They were not allowed to make any profit, right? Now that's been changed and broken. Like here, Anthropic aren't saying we're not to make any profit. They're saying we're absolutely going to make some profit. However, you know, with the guardrails of making sure we are responsible to the human race, by the way, it's very much within the company's interests that they don't break the planet and destroy the human race, right? So I don't see this as a contentious thing. Obviously, you're going to make decisions about stuff like when are we going to release our new latest model that's going to be better than open AIs, you know, in that competitive arena they're operating in, when are they going to release models?

7:55Piers Curran: Are they going to maybe delay the release? Well, of course they are, if it's not safe to release. And so I don't see an issue here. This is, I think everyone's aligned. Yes, they're going to make a load of money without breaking things along the way. It kind of feels a little bit like everyone's caught up in the AI bubble, so unwilling to look beyond the superficial. When you think about other technology matured companies, which are more product orientated, like Apple, for example, everyone still buys Apple phones. I, in fact, upgraded to the 18 because I had a 13. And so the point being, though, that we know that the innovation is probably far further down the track than what we're receiving in the 18 pro but they stagger it in a way to just draw it out because they know that that's the most assured smoothest path of revenue over time so similar here isn't it it's not that we're gonna halt i guess the naive way of looking at it is that if you delay the latest model then that's going to impact short-term revenues but in reality and And we'll go on to discuss, is part of the influence of Dario and others to shape the regulatory environment already going to help control the moat around their technology to that respect?

9:17Piers Curran: Well, exactly. And Apple can do what they do because of their moat. So you can only do that if you have a moat. A moat being, I've got a better product than everyone else. and or because of that over time i'm so integrated into my customers processes and behaviors that i can afford to just stagger my releases um but unfortunately i think we're still early days in this race i mean i know you know it's only what is it like four years since open ai kind of dropped their chat gpt b2c facing and the whole world went mental we're only four years in here so this is still at the start of the race um so i don't think they've got a moat like apple have um is my point and so but they do have a responsibility to themselves and their shareholders and by default humankind to not put out models too fast that aren't safe i i think this is fully aligned they're commercial they are commercially incentivized to not release models that aren't safe as well as obviously being morally incentivized to do that so let's talk about just top level the different classifications of the share classes and then what is the purpose of each one in simple form so i mean i said earlier the purest and like if we just talk tech and go down the years so apple and amazon who ipo'd many moons ago they kept it simple one share equals one vote that's it right i own the company proportionately i get a say in how it's run meta and alphabet came along and did things differently these are the founder led you know tech giants and so for them the founders hoarded basically high vote share they had this share classification that was just voting shares right and so they get to control so zuckerberg has 50.1 percent uh or maybe he has more anyway he's got the majority of the voting shares in meta he singularly as an individual decides what that company does um even though he doesn't own 50 plus percent of the shares he owns the voting right shares that gives him the majority on the voting side.

11:42Piers Curran: So at the board level, he can make or veto kind of decisions, right? Now, anthropic, basically, they're totally separating everything out. So you're going to have different share classifications for, let's say, the money, right? Do I commercially benefit from the growth of this business if I'm a shareholder? Great. That's the money side, the commercial side. There's the voting part. So, right, what say do I have in the strategy in the direction of a business. And then there's the board control, which is the complex piece here. So the board control, you know, well, we'll talk about it now, but there's these kind of share classifications that try and cover off all of these, right?

12:23Piers Curran: So class A, that's the economic part, right? So this is the public. So if you're going to get involved in this IPO, and you're going to buy some shares in this IPO, you're going to get class A shares. and this gives you zero power so you don't get any voting rights but you proportionately economically benefit as the company grows or you lose out if the company blows up right then we have the class n shares this is pure profit this is zero voting rights right so sidelined um or yeah i mean i guess this the reason so amazon we'll get into the circular nature of all of this some big shareholders who own the business now before the ipo it's amazon and google are massive shareholders in fact amazon if they convert some of their their notes will actually own 20 of this business which is like kind of insane right but what's really interesting here they've got a class of share just for those guys it's class n so they'll again benefit economically but no voting rights and the the main theme here is to not trigger any anti-trust alarm bells so these big tech will not have a control over anthropic and how it's run to satisfy the anti-trust part right and they won't have a steer on the product roadmap either um so to keep it all separate, but they're still in the game.

13:57Piers Curran: They've got skin in the game in terms of benefiting economically. Then you've got the founders. So here you've got your class F founders shares. These are basically each founder gets a single golden share and that automatically locks in. Here's where I got my 50.1%. So this locks in 50.1 % of the shareholder vote. So the seven founders get a golden share, giving them the majority. It's interesting with the founders though there's seven of them you know what's their relationship like i'm sure it's tight at the moment but will they fall out because it only takes one it only takes one founder yeah to get off the the script if you like and then all of a sudden they don't have the majority of the voting shares which is an interesting that's going to ratchet up a lot once you're a listed company and you have to report every quarter that that that tension between those seven yes it's interesting dario in the in the filing here in the s1 filing basically put as one of their strengths was that all of the founders are still at the business running the business which was basically a side dig swipe um at sam altman because there are no founders of open ai still there apart from altman himself well do you know do you know where these seven founders of anthropic do you know where they came from right what were they doing before anthropic oh i can't remember now i did read this well they were the competitor now working for open ai yeah they span out of open ai oh because they didn't agree with sam's yeah the yeah the whole ethics of it exactly that's right right so they span out and set up on their own just five years ago.

15:47Piers Curran: I'm going to talk some revenue figures in a minute. Yeah, I don't know. This company didn't exist. This company did not exist five years ago. Just unbelievable. Anyway, so you've got your founder's share classification, your class F. Then you've got, then it gets complicated with regards to the board. Because remember, with a company structure, it's the board that basically control and govern the company and its strategy and its direction right so who sits on the board is really really critical so here they have a what they're calling a class t share um so these these have no economic rights whatsoever but it's basically Basically, it's an independent panel of non-investor safety trustees.

16:39Piers Curran: So they're not economically incentivized at all, but they control, essentially, in some ways, they control the board because it's this group of trustees that have control of who sits in four of the seven board seats. so the trustees control the majority of the board in that they get to pick who sits there. Okay? So that's your class T. Can I just say, I just want to put out there now, I was just having a quick look, who could these potential people be? Because they are subject to bribery, baby. Well, do you know who they are? I don't know. I know two of them. Well, I know two names. To be honest, I didn't know one of the names, but there's a great name I'm about to say who's on this committee.

17:35No, it's not. This episode is brought to you by Indeed. You're ready to level up your business, but the pressure to find the right person to make that happen is intense. Indeed Sponsored Jobs can help carry the load. It boosts your job post to help it reach quality talent fast. Cut through the chaos with Indeed Sponsored Jobs. Get a$75 sponsored job credit at indeed.com slash podcast. Terms and conditions apply. Hiring now? Then this is a job for Indeed Sponsored Jobs. Mr.

18:07Piers Curran: Ben Bernanke. Okay, yeah, yeah. I mean, I've got a list here, but this hasn't been... Fabius Fed Chair, the Federal Reserve Chair during the financial crisis, Mr. Ben Bernanke is on this Board of Trustees, which is interesting. I've got a list here. Yeah, Ben Bernanke. Richard Fontaine National Security Expert Head of the Center for New American Security Neil Buddy Shah CEO of Clinton Health Access Initiative and former Agenting Director of GiveWell again this hasn't been verified but Christopher Stubbs a Harvard physicist yeah just just interesting like I wonder what the relationships are I mean obviously we know Bernanke very well but the others less so to yeah I suppose Bernanke's probably got a track record where being in the seat that he was as Fed chair, you would like to think he is as independent as you could be because that's exactly the role he facilitated.

19:06He's essentially the most important person on the planet for that period in 08, 09.

19:11Piers Curran: A safe pair of hands, right? A trusted steward, perhaps is the best way to say it. So fine, right? So you've got this class T, then you've got class E shares, final class T, This is actually for employees, but it's actually, this part is that they're non-economic shares. The employees do have other share classes, so they do economically benefit, but there's a specific share class for non-economic veto shares. So this comes back to who sits on the board, right? Because you could say, well, who picks the board seats controls the company. So here's where it gets a bit complicated. and maybe I can just try and go through it.

19:53Piers Curran: There's basically two power centers in this business. It's the company's seven founders and then it's the five trustees, okay? So it's the trustee's job to preserve the balance between the financial interests of shareholders and then the company's non-financial purpose, basically having AI be good for humanity, right? so these trustees as i've said they get the right to nominate four of the seven board directors the board in turn then makes the big corporate decisions and strategy right voting shares decide who gets the remaining three seats so basically here's your 50.1 percent founders control so Essentially, the founders pick the other three.

20:45Piers Curran: But then you're like, okay, that's interesting. The founders are basically giving control to the trustees here, essentially. But there's an important kind of side point. There's another, what's called a nominating committee. So you've got your founders, you've got your trustees. Founders get three seats, trustees get four seats. But then you have a nominating committee, which has three members, including Dario, and they can basically veto the trustees pick for the fourth swing seat. This committee, they can't propose their own candidate. They can just block it. So I guess then the committee, the trustees have to go back and find someone else, right?

21:38Piers Curran: So there is this little bit of a safety net here for the founders in that the trustees can't just pick four they want. They can pick three they want, and that swing seat can get blocked. So it's just, I don't know, I mean, why? Who comes up with this stuff? Yeah, I mean, I'll tell you who comes up with this stuff. There's 22 banks on the ticket for this deal who've been beavering away with the lawyers to come up and engineer this solution. the lead left so the kind of golden prize for running this IPO is Goldman Sachs followed by Morgan Stanley but as I said they're all on this given the size and scope of it.

22:17Question then given what you've just described there so if Dario and Wall Street are the hunger for commercial revenue once you're a listed entity but the trust decides whoa whoa whoa guys this is way too risky for humanity in whatever shape or form that might be. So the trust wins is my understanding. So as an institutional investor then, thinking about at some point, this is going to have to go into people's portfolios and so on, analysts are going to have to value this company. What multiple do you pay for equity when the board's legal fiduciary duty is explicitly not to maximize the share price?

23:01Piers Curran: Yeah. it's a very good question i i look at it i go back to my point earlier to be honest i like if you try and a lot of people are overthinking this i believe you know as i've said it is not in your moral or commercial interest to release a model too fast or open ai have released their latest one is better than ours quick let's release our next gen one to get back in front well you're only going to do that if the next gen one is safe if you if they release something that's not safe that would be not only dangerous but the biggest own goal commercially because if you lose trust from your users well then ultimately you're dead because there are alternative products in the market or just go to open ai then we'll go and use gemini right i mean it's annoying because it might be embedded in your processes and stuff but ultimately if you can't trust the supplier you go somewhere else right so i i don't think investors are going to overthink this it's fine they're banging their drama about oh we're going to end the world if we're not too careful well sure um but i think the commercial side and the moral obligation are very aligned here and probably the natural life cycle of a startup if you can even call these types of companies a startup is that it goes through this exponential growth phase but at the point of listing you then do have these fiduciary duties these regulatory requirements and so forth that slows the pace of innovation anyway and then your company infrastructure i all the boring nuts some bolts of what make a larger organization function, which have not been developed because of the acceleration that was tech first.

24:54So Anthropic, in the end, what will be interesting is because I was reading an article this morning and it was talking about how a lot of hedge funds are sponsoring and paying PhD candidates. So the likes of XTX or Citadel, G Research, I think QRT are paying out for 70 PhDs at the moment, hoping that one lands kind of VC style because that will pay for all. But yeah, I wonder if it felt like all of the talent was getting hoovered into Anthropic over the last three years. I wonder whether as that pace of innovation starts to slow naturally, whether the talent pool, which the hedge funds have struggled with, starts to then come back a little bit.

25:42Yeah. Because they do have the firepower as well from an income to invest in technology perspective. And we know those hedge funds are willing to push the boat when it comes to innovation because they are profit seeking. Yeah. Yeah.

25:58Piers Curran: It's true. And I think that, you know, this, um, this moral bandwagon, I think there's two parts to it. It's definitely to attract, it's a strategy to attract talent. And look, we are on a mission to save humanity and, but we've got to be super careful and it's risky and dangerous, but you know, we are at the frontier, you know, come with us, we're going to change the world. You know, that's a great way of attracting the talent, right? But yeah, as it becomes more commercial, That line kind of gets diluted a little bit. But then on the regulatory side, Anthropic themselves are banging the drum the loudest about the dangers of AI.

26:39Piers Curran: Now, again, I put my skeptic's hat on. Is this a clear-cut strategy to protect their moat by basically forcing governments to implement regulation? Regulation slows everything down. It's actually anti-competitive. But look, the dangers are so great, we need regulation. so that it's kind of a way of them kind of protecting their moat by forcing the governments to regulate. I'm like, well, look, it's your product. You know, you shouldn't be sat there relying on the government to make your product safe. That's your job. That's your responsibility. Don't relief stuff that's not safe. So this whole regulatory thing I think is a bit of a hoax where they're just trying to protect their moat.

27:27Yeah, so to decode that, that's a strict regulatory regime raises the barrier to entry, and that would then penalize under-governed open source or smaller competitors while favoring Anthropics compliance infrastructure we've just discussed. Yeah. Okay, well, look, we'll circle back to the risk factors in a moment. let's have a little look at the actual financials you know this is one of the times when you know they do start to i think as steven says lift the kimono and we get a little look at some of these numbers so what do those numbers look like and you having computed them how do you feel about them

28:12Piers Curran: i feel pretty good about them the market feels really good about them that's why i said at the start, we've had this kind of broad-based, I would say, AI trade uptick since they've released this S1. The numbers are staggering. We kind of knew that on the revenue side, right? So, and here's, yeah, like if we take H1 of 2025, then they did$1.1 billion of revenue. The same half, you know, H1 2026, so that's obviously January through to end of June 2026, the same period a year later, that went from 1.1 billion to 16.3 billion, which is 14x year-on-year growth. I mean, look, companies have grown. I'm sure you can pick an obscure tiny little company somewhere that's grown faster than that.

29:01But at this scale, that is a world record.

29:07Piers Curran: You cannot find me any company ever in the history of mankind that has grown like that, at that scale. I would say I feel like Anthropic, given the way that the broader market has responded since the filing hit, which is in a positive fashion, given some of the anxieties that were being exhibited in the market. Yeah. They definitely did a good, their spin doctor did a good job in the week before they put this out. Because you remember the FT exclusive and they were leaking source comments. Oh, actually, it's not been quite as rapid as we were saying initially. and then this lands and now you're saying it's great 14x it is bar suitably the problem with the problem they have is they've kind of set their own bar which is like the highest bar in the history of mankind 14x so now investors are like well okay what's the price of the shares please that you're asking me to pay and then what's your growth like in the future because as an investor in the IPO, you're buying the future, right?

30:10Piers Curran: So the revenue growth has slowed. So I think in Q2, it doubled versus Q1. But now into Q3, it looks like it's going to be more like 50 % growth. This is quarter on quarter growth. It's still kind of insanely fast, but the rate is inevitably slowing, I guess. But then like, it's not just about revenue, of course. It's, well, okay, is this company going to make any profit at any point? Because even though those revenue numbers are staggering, they still lose money. So we look at revenue per dollar of compute. When we, you know, we're obviously thinking about, well, how much are they spending, you know, versus their revenue?

30:53Piers Curran: Are they going to turn a profit at any point? So for them, compute is like their biggest cost, right? So they're having to pay for compute power. Then they're selling their product to their clients. And how much are they charging their clients? And is there a profit margin there or not? Okay. So we look at revenue per dollar of compute. That's really, really improved. I think this is one of the best parts of the S1, to be honest. so basically even though they've obviously accelerated so research and product velocity and they've expanded internal usage of their own stuff so they're using clawed a lot more internally which has sped things up so basically their revenue you know grew 1 368 percent year over year as i've mentioned there that 14x but the compute and infrastructure expenses have grown only 300 only 300 right so so obviously revenue's growing at a much much much faster rate than their main cost base so it as long as you extrapolate those trends further into the future well then of course at some point you get that kind of tip into profitability okay um well one of one of you know one of the things that well maybe what we'll do is we'll flash a chart up because one graphic that I saw circulating on X that a lot of the people in the know were looking at was exactly that significant increase in revenue per dollar of compute.

32:28And you can see on the chart, which we'll share, is 5.2x increase in compute efficiency from Q1.25 to Q2.26. And you can see that pickup that's been pretty consistent now over that period. but one of the things i remember we were talking about with core weave and core weave's been in the crosshairs in the last fortnight yeah as as part of the yeah anxiety around this whole ai funding circularity and so forth has come into question something that we did see was the concentration of some of these base customers and it kind of reminded me like when we were looking at i remember I think it was a filing that we had for CoreWeave and they had quite cool graphics of what customers, but not just what customers, what contract lengths and what contract sizes were they tied into.

33:24Because obviously longer contracts all locked in, that's generally more healthy, a better situation rather than the complete opposite would be spot price contracts, which apparently when we were talking a week ago before this filing came out, um spacex were placing huge spot contracts with anthropic you remember at the time there's a little bit i scratch your back you scratch mine going on but that's not a good thing because it's

33:52Piers Curran: non-repeatable in that sense so what did we fix your cost yeah what what did we get them as on that side on the customer base and what they're spending yeah so look there is a concentration risk um but it's no nothing like core weaves i think what we say in core we've 75 of their revenue was from three customers so here for anthropic is 25 of our revenue comes from five customers okay yes that is a concentration but you know these customers are the biggest yeah beasts on the planet, right? And so they're good for their dollar. You know what I'm saying? So hang about, these companies, for context, have got a higher credit rating than the US country.

34:41Correct.

34:44Piers Curran: So obviously there's that, right? But I don't see 25 % concentration of hyperscalers. I don't see that as a risk. But then they split it out. What was a bit more interesting was, well, how many customers are paying? So they split it into different categories. So customers that pay more than$10 million per year to Anthropic, customers that pay more than$1 million, and then customers that pay more than$100 ,000. Now, as you'd expect, up at the$10 million plus, you've only got 250 companies sat at that run. Companies, clients, customers that pay more than a million dollars a year, that's 1 ,900 customers.

35:23Piers Curran: And their customers paying more than$100 ,000 per year, that's 11 ,900. So it was interesting just to see the spread. Because look, if you want to go from 16 billion now to 100 billion of revenue, which is what they're planning, you know, some talk about a trillion dollars of revenue a bit further down the track. But to get To get there, right, to get from 16 to 100, to add 84 billion, you're going to need some really big clients spending loads of money, right? It's much easier to onboard new logos that are big spenders rather than go out and try and onboard a million small logos that are spending a tiny amount, right?

36:10Piers Curran: So how many companies are there that can spend more than$10 million a year on their ai spend i don't know yeah i mean i guess i would be slightly concerned if that was just like a any other normal product but because we're talking about the exponential growth of technology that we've had through ai it's kind of well how big could google be well it could continue growing and growing and growing and growing it's not like a saturation point uh of course there's some risk But yeah, I mean, if you're looking at those numbers, you said if, what was it, nearly 12 ,000 customers, the overall majority, that only accounts to 1.19 billion in revenue.

36:55Yeah. You said that they're clocking 16. So that's right. Just goes to show the magnitude of spend that 250 at more than 10 mil. I'm sure that skews massively on the tail for multi-billion spend by a couple of those five firms.

37:10Piers Curran: yeah but look i mean i don't know how many companies there are on the planet i mean what's their total addressable market their tam and i mean i don't know ask ask claude how many companies are there on the planet like millions millions millions millions right and if they get this right every company on the planet should use their tools so there are millions and millions and millions and millions of small clients and customers that they can hoover up. It's just that might take a bit longer than hoovering up big, big clients with massive checkbooks. Another very interesting story I saw as well, just while all this was going on, is the FT reported that Anthropic are investigating.

37:57I think they've maybe even acquired a small European consulting firm. This episode is brought to you by Indeed. You're ready to level up your business, but the pressure to find the right person to make that happen is intense. Indeed Sponsored Jobs can help carry the load. It boosts your job post to help it reach quality talent fast. Cut through the chaos with Indeed Sponsored Jobs. Get a$75 sponsored job credit at indeed.com slash podcast. Terms and conditions apply. Hiring now? Then this is a job for Indeed Sponsored Jobs. I was talking about this with Stephen. We've got an episode coming out next week where we talk about, is AI going to kill management consulting as an industry?

38:41Piers Curran: Oh, interesting. And what I said to Stephen at the time was, I don't understand why Anthropic don't just start their own consultancy firm. Because, and since your share price has gone down and down and down and down, it's like take some of the greatest talent within that firm. And now that you're a commercial entity listed firm, and why don't you just embed your consultants who are anthropic consultants, i.e. second skin from Accenture or McKinsey, deploy them into your five or to your, how many was it, of your 200, your biggest spend? So take your more than a million dollar spend. So let's say I hire two and a half thousand consultants from all the big ones, BCG, McKinsey, Accenture, hoover them all up, pay them more money, and then say, you need to make sure that all of these companies, our technology is absolutely entrenched in every conceivable workflow that they have rather than outsource it.

39:37And then write the research telling you about how much more that you need in the future and the opportunity ahead. So they've already started this.

39:44Piers Curran: But maybe bad news for McKinsey, but is it bad news for the consultancy sector? Well, it depends what job level maybe you're looking at. Yeah. True. Yeah, no, it's an interesting one. But let's talk about then the other parts of this then on the loss. I mean, there was... Well, the loss. The loss is interesting. I mean, look, the revenue is spectacular, right? But as we said, compute ain't cheap. NVIDIA chips, they're not cheap, right? So what they're having to buy, they're spending more than they earn, right? So they're losing money. Now, the alarming headline out of the S1 is that the filing shows an eye-watering$80.7 billion net loss for the first half of 2026, which on the face of it is like, bang, whoa, they are so far away from profitability.

40:40Piers Curran: Like on the surface, you're like, what? But the key, key, key part of this is the absolute vast majority of that loss in inverted commas is what's called a non-cash accounting charge. But in a weird way, the bigger that loss, the better. Because actually that loss stems from the likes of Amazon who own convertible notes. And so these are products that convert into shares. But at the time these convertible notes were issued, you know, the valuation of the business Anthropic was way, way, way lower. right so but you've got to value that convertible note at that point of issuance this this this non-cash accounting charge is the gap between well what was the value of those convertible notes at issuance compared to what's the value of those convertible notes now if they were converted into shares at the ipo at let's say a two trillion dollar valuation so it's just the gap in the valuation of those notes.

41:40Piers Curran: And it just so happens to be$72 billion of the$80 billion net loss. It's just an accounting charge because the value of their company's gone through the roof. It's not, they don't have to pay$72 billion, right? They don't owe that money. That is just a valuation change in the convertible notes that Amazon are going to be laughing about because where, hey, you know, we're going to convert, obviously, these notes into shares, and we're going to realize a huge gain because the valuation of the business has gone up. So my point is, the bigger that loss, the better, meaning the higher the valuation of the overall business, right?

42:18Piers Curran: So if you kind of take that off the table, basically, they spent nearly$9 billion on compute and infrastructure in the first six months of 2026. And basically, that winds up as being their loss, right? That's essentially. And one of those charts that I did see was Anthropic spending on compute power is increasing, but its revenue is growing faster. We'll put it on the screen. You can see that revenue, like the gray bar, at what, 11.6 in Q2 2026. That's just jumped massively, isn't it? And the gap between that, if you look, it switched at what the beginning of this year is when revenue started to overtake compute and that's probably what you would have expect with the creation of a business of this type right that's right exactly this this is a good news chart right this is this business model is working you know it's taken it hasn't taken long at all they've only been alive five years but it's like it's just slowly each quarter that revenue bar is just eating into the the kind of compute and infrastructure cost bar and now suddenly the revenue shot off and you're really seeing the return on investment which is that what everyone's worried about all of this mega spend on ai is there a return on investment at the end of the day and this chart says yes there is and i think this is one of those good news pictures that's helped with that general vibe you know around the whole ai trade over the last couple of days and maybe just on the contract side we briefly touched on it but the filing did have some color about the relationship with these hyperscalers yeah the well i mean so one of the concerns then so obviously that was a good news shot one of the concerns people have is the commitment that anthropic have made to spending on this stuff So it's like a, it's a 500 to$750 billion of future spending commitments that essentially is non, it can't be canceled.

44:33Piers Curran: So they have committed, let me say that again, they have, let's go top end of the range. They've committed to spend$750 billion over the next whatever decade or so to continue this roadshow or this circus or whatever you want to call it. But as long as the revenue is there, as long as the revenue growth continues, then absolutely spend this money because the net return on the dollar invested is positive. your profit margins there right so like and this is where it gets a bit circular though because i've mentioned that google and amazon big shareholders in anthropic but they're also their biggest suppliers so out of this 750 billion spend they've committed to 111 billion of that is going to google cloud 110 billion is going to aws amazon cloud service you've got microsoft that are in line to get 31 billion right so you've got these uncancellable commitments to compute which they have to do to keep the growth story going and as long as there's a customer at the end of the day who's paying for this stuff and that revenue growth continues it's fine it's just that that revenue growth has to continue is this the case of you keep your friends close and your enemies closer.

46:04And what I mean by that is, you know, when I think about Amazon and Google, so when I mean close, I mean really close, like equity, let's make them class and shareholders within our business, like literally. But, you know, when you think about Titan, Olympus or Gemini, so some of these other products, the latter you probably heard of with Google, but other with amazon bedrock vertex ai all these other ones and then the s1 explicitly warns that these hyperscalers could curtail delay or reprice compute allocations or prioritize their own internal ai labs over anthroponic yeah so is this i mean you read that and you go oh my god like it's It's quite scary, but actually just being rational about it, isn't this just sensible?

46:57This is just a sensible commitment to make sure that... It's almost like the US and China making some trade agreement and they have to put up some collateral, some commitment collateral, in order for both the agreement to function. Even though we all like each other, it might always remain the same.

47:16Piers Curran: But this wouldn't have been possible... without the hyperscalers balance sheets you know to lean on because the amount of money that's required to get this show on the road this ai you know data center build out and you know you you couldn't have done it without the hyperscaler balance sheets so it's not a case of oh isn't this all circular incestuous it's the only way it could have worked and and yes it is a little bit circular and incestuous but that that was the only route so getting this stuff out to the masses you know quickly so but you're right there are some conflicts of interest um but that's what you got to put up with um because without those that that's kind of the inevitable by-product, if you like, or consequence of the only pathway of getting this show on the road.

48:20I did see some notes that you'd made about the case for software front end. And it was looking at the economics of moving like software in itself. What do you mean by that?

48:31Piers Curran: Yeah. Well, it was just like, how do you think about Anthropic as a business? This comes down to, well, how do you value it? You know, what should I pay for the shares? You know, and is it a software company um and and so if you think it is well this is where you're thinking about okay anthropics spent so here's a good example of of their spend right so they spent two dollars fifty on compute for every one dollar of revenue in 2025 okay so two dollars fifty for every one dollar of revenue because of the revenue growth that equation's now flipped they now spent Instead of$2.50 for every$1 of revenue, obviously losing loads of money, they now spend 50 cents for every$1 of revenue.

49:21Piers Curran: Okay, so that direction of travel is big, right? And then when you're thinking about like Q3, like thinking into the future now, I know Q3 is about to end in two days' time, but Q3, well, sorry, Q3 has already ended, what I'm talking about, end of September. But in Q3, they reportedly got one, they earned$1.70 for every$1 spent on computing power. So that profitable, that operating profit margin is now visible. I think another big good news story here of this S1. So they posted an 8 % operating margin in Q2. And that was compared to a$1.2 billion loss a year earlier, right? So the tipping point has happened, I would say.

50:03Piers Curran: But is this a software company? you know you're looking at revenue growth um and that has to outrun your cost base okay and that's like kind of the software model and we've seen it okay um but then what are your commitments i go back to that spend commitment are they fixed are those fixed costs and the problem that anthropic have is no they're not and then you look at the spacex deal where they're leasing some compute from elon but it's on a 90-day cancelable contract and elon knows that the cost of compute is going up and up and up and up and up and so elon's like well i'm only going to commit to 90 days because in 90 days time the cost of compute will be higher let's redo the deal and pay me more right so it's not quite as simple as to say the thick the costs are fixed for anthropic into the future so it's just about revenue growth and your margin popping you know so we'll see but obviously it's all about the revenue growth and if that stalls well you're in big trouble the house of cards collapses so you go back to the revenue growth and that is demand what is the demand story for your product and i think it's fair to say the demand story is just still wildly positive we're only just beginning in terms of adopting AI, right?

51:26Piers Curran: And so I think as long as you're okay with that thought, then fine, this revenue growth thing is going to continue. All right. Well, look, we've got two more areas we need to cover. One is the valuation. I want to talk a little bit more about how you think that looks, but the kind of process and mechanics behind what's going to happen now over the coming weeks, and then the actual listing in itself. And then of course, the risk factors, which we've touched upon throughout, but maybe just to consolidate thinking around that part. So the valuation target we mentioned earlier, there's like 20 plus underwriters syndicates on this particular deal.

52:01And they're looking at an implied valuation around one and a half to two trillion. Remember, that has come down again. The bankers have done a little bit of the dark arts here where we kind of raced up to two trillion and we've just toned it down a little bit just to ensure this gets away nicely but yeah what does the valuation multiples look like here well at two

52:24Piers Curran: trillion so that's that's this kind of very wide i mean think about the range 1.5 to 2 trillion that's a 500 billion dollar range right so it's kind of all a bit nuts but if you think about 2 trillion, then that would be basically 20 to 28 times forward 12-month gross revenue. So what we forecast Anthropics revenue to be over the next 12 months multiplied by 28. So you're paying, as an investor, you would be paying for 28 years worth of revenue, not profit, 28 years worth of revenue, but that's assuming there's no growth, right? So obviously we're expecting spectacular revenue growth to continue, which is why it can command such a crazy revenue, forward-looking revenue multiple, because that is a crazy multiple, right?

53:26Piers Curran: So we'll see. One and a half to two trillion. It's so hard to kind of price this given the very unique risks that we've already talked about. If I was Anthropic here, I mean, they're not going to listen to me, obviously. But look, there's bumps, there's speed bumps in the road. Okay, we've already had a big wobble on the AI trade over the last few months. It's not going to be the only one. This is going to be a volatile thing. If I was them, I would just lock in my funding for the next few years right now at this IPO. I would just go, let's just raise$250 billion. Now, today, let's just get it in the bank.

54:10Piers Curran: This will help us ride the noise and the volatility of this AI boom, and we're locked in and we're safe. So that's what I would do. So I wouldn't be surprised if you get a valuation at the bottom of this range. Because to raise that amount of money, you need a lower valuation to attract in more buyers, right? So I wouldn't be surprised if you get a one and a half trillion valuation, but they lock in huge sums, like 250 billion. So it depends, right? But if they don't want 250 billion, if they only want 100 billion, well, fine, go to 2 trillion. You can sell that. No problem. So I guess this is a slightly different problem that Elon had with SpaceX.

54:56SpaceX could have raised more, but he kept it pretty tight and narrow. so demand was going crazy at the time multiple times over, oversubscribed so you're saying the opposite let's just get the funding in because of the nature of compute power and spend and then we just assure the years ahead that's what I would do

55:15Piers Curran: but obviously they won't listen to me I don't know, it'd be really interesting I would just lock it because it doesn't matter to them I mean the fame does now one and a half trillion, two trillion I mean it's so ridiculous it's the same right so I would be forward planning my business through some volatile months and years ahead let's just lock it in now let's just remove that risk I do like your rationale there just given I feel like we've have got a few steps closer to a little bit of a correction coming in regards to the AI not bubble bursting no way but some kind of like realistic recalibration of reality once these companies kind of get to maturity which is listed then you know whether that happens post-anthropic or post-anthropic and open ai in early next year i don't know but it feels like it's almost a necessary function for the second chapter of these companies to go through having gone through the infancy and growth so now the more maturity stage one thing was lockups because spacex had five release dates in the first seven to six days and you know prior episodes we talked about how this is a strategy to try and just control uh the nature of so many force heavy selling that might come into the share price any thoughts on that on yeah i think that i think spacex uh they did it they did it well i mean we definitely got some downside on the spacex you got a pop higher um in the days following the IPO on SpaceX, then it trended down and we broke below the, you know, we hit$100, we broke below the IPO price.

57:04Piers Curran: So I think you've got to spread it, which I think they'll do, like from a founder's point of view, you know, so that's employees, right? I think from a founder's point of view, it's normally like 180 days, isn't it? It's normally a 180 day lockup, meaning founders can't sell their shares for the first six months, essentially. So I think they'll kind to do similar um but yeah it matters for the float doesn't it but this is back to my point if they raise a load if they do go big and raise 250 well the float's much bigger um and which can then absorb better they're like it kind of yeah it shelters or dampens the impact of employees selling their shares if the float's much bigger.

57:54All right, well, let's close and finish with the risk factors. And one of the things I was thinking, all the conversation we've talked about is the ever-increasing demand side of this. And we just kind of talk about compute in quite a nonchalant way. It's like, we need more compute. However, it doesn't just happen like that because there's a physical limitation to obviously the power that's needed so you know electrical grid interconnection delays nuclear clean energy you know these are things that could be national security massive speed bumps along the way to selling a frontier model yeah in in this sense on a on an enterprise commercial basis feels like uh yeah the relationship that this company will have with the US government, Trump or Trump aside, because they won't be here long, well, touch wood, they'll be here long after Trump has then left office.

58:55Yeah. But it feels like we're moving into a new political, geopolitical dimension where one of your most strategic and most important assets is a company like this. Yeah. In its enablement of economic prosperity through productivity and how then these relationships are going to play out in the Western world against, say, China and others. Yeah. It's getting quite fascinating.

59:26Piers Curran: I think there's a big political risk for Anthropic. Because like if in the US at the moment it's become Republican versus Democrats the AI, where you sit on the whole AI story has become a very political thing, right? And so Democrats are very much about safety concerns. Let's halt AI progress because of the dangers and the risks. Let's block data center builds because it's disrupting and having a negative impact on communities. And all of that kind of language is very Democrat. The Republicans are trying to have the opposite voice. Let's go. this is amazing for the economy let's build as many data centers as we can you know so if the democrats so the midterms next week you know if the dem attack if the demos take control and and does that kind of pave the way to a higher likelihood of a democratic president in a couple of years time that could slow things down because i think ultimately that's out of their control to an extent right anthropic you know the regulators will dictate the speed speed of data center build out the speed at which we can increase electrons right power you know this is like very much sits at a government level and i think ultimately that's going to be the controlling force on the speed of anthropics ability to grow so interestingly i had a a source that i met at the weekend oh yeah he works at one of the leading u.s tech firms and sits very close to the top table so has an ear on what goes on and there was a meeting two weeks ago that took place with with potus right and everyone was there including anthropic as well as all the other big hitters elon so on and they were talking about this regulatory stuff and uh one of the ceos was talking about prior experience and the experience that they had had in Europe and how the European regulatory environment is very different, obviously, to the US.

1:01:44And the advice then to the president was quite clearly that we need to pivot the optics terminology built around in the public domain of AI to superintelligence. Yeah. Interestingly - Rebrand it. But interestingly, the day after or two days after, and I was like, okay, yeah, didn't think anything of it. SpaceX released their earnings a week and a half ago, and they renamed their XAI division to superintelligence. As to have a number of the other leading tech firms have started to refer to the broader name as superintelligence. and it only gets referred to as AI in certain context. Yeah. And that's all to do with, yeah, as you've been just rightly talking about, the regulatory environment that we're heading into.

1:02:41Yeah.

1:02:42Piers Curran: This rebrand, like the PR campaign for the AI boom has been an absolute disaster. The stats in America are like, there's a vast majority of Americans are anti-AI because of how the media has fueled and fueled the flames of all of this doom risk and, you know, it's going to kill humanity. And, you know, so there's a PR exercise required. I think the super intelligence thing is the start. I think the horse has bolted. I mean, can you rebrand it now? I mean, I feel like it's kind of too late. I mean, who's going to start saying SI instead of AI? I mean, I don't know. Anyway, at least they're trying.

1:03:32Piers Curran: But look, for anthropic, to a degree, its growth potential is out of its hands. And it's not from the demand side, it's from the supply side. It's the supply of compute and power to meet demand. And I think that's where the biggest problem lies. Just got to share that money around the circle, Piers. Help a brother out here. So in conclusion, you feel quite positive then from what we've seen so far? I think this is good. I think it's just reignited the AI trade a little bit. Look, we're only a couple of days in. I think it'd be really interesting. I'm not too bothered about the valuation. You know, if the valuation ends up being 2 trillion or 1.5 trillion, I don't think it really matters.

1:04:27Piers Curran: If I was them, I'd go lower to lock in a larger amount of funding to kind of smooth the journey in the years ahead. But look, ultimately, they're on a pathway to profitability. And as long as the revenue keeps coming, baby, and as long as they can supply their product to their customer base, then look, I think this is good. AI Trace back on. What bubble?

1:05:03There we go. I'm going to clip that last little phrase when we hit 10 ,000 in the S &P. All right, cool. Let us know your thoughts, how you felt or feel or read and interpreted the S1. We'd love to get your opinion. Drop us a comment. If you've made it this far and you're not subscribed to the show, remember to hit that subscribe button. More episodes coming every week. Thank you, Piers. Thanks a lot.